I spend a lot of time talking about asset allocation on the blog. I’ve presented quant systems that focus on price based tactical asset allocation, cross-market momentum strategies, and mean reversion strategies (just to name a few). Some older posts are:
A Quantitative Approach To Tactical Asset Allocation
Mean Reversion After Really Bad Months
Asset Class Returns Based on Fed Policy
Inflation and Asset Class Returns
I was originally going to put this out as a white paper, but seeing that I have about 5 of those on the back burner (and this seems to be a pretty well known property), I thought a simple blog post would do.
Below we examine the effects of the yield curve on our group of asset classes.
(Data sources: Global Financial Data)
US Stocks – S&P 500
Foreign Stocks – MSCI EAFE
Bonds – 30 Year US Govt
Commodities – GSCI
REITs – NAREIT
Spot Gold
Buy and Hold is an equally-weighted, monthly rebalanced allocation to the above 6 asset classes.
First, we examine how the 10 year US Govt Bond – 90 Day Tbill rate affects the return of various asset classes. We selected three modes, but by no means are these optimized or optimal. The table below presents the % of time spent in each mode, as well as the annualized returns for the annualized returns for the asset classes.
One could devise a simple timing system based on these properties.
When the yield curve is <0%, long commodities and gold.
When the yield curve is 0 – 2%, long us and foreign stocks.
When the yield curve is >2%, long bonds and REITs.
This simple system would have beaten buy and hold by over 4% a year over the time period with more volatility (mostly upside) and a similar drawdown. Equity curve below:
And since you asked, here’s the chart with the momentum rotation system added with a simpler combo of both too:
be interesting to see results trading HY bonds
For peroids when the spread is >2%, it would appear, based on table, that EAFE would be better than bonds. No? In fact, it appears that bonds are never in the Top 2 performing asset classes.
Similarly to trickymick's post, why not commodities and US Stocks when 0-2%??
I would like to see the results using 10 year corporates minus 90 day commercial paper. I think this might lower the drawdowns as Gov paper reacts differently than corporate in times of stress.
First, thank you for the education, here, in your white papers, and in Ivy!! I am a CFP, ChFC, trying to educate myself about alternate investments and way to protect on the downside (I was in cash for a lot of last year until the “fall”, but know I need more discipline in my investment approach). I finally have the time to work through a trial on AlphaClone.
Question: Will your bond assumptions still work in a secular bond bear market? Is the 2% as valid for interest rates very low and for ones much higher? A 2% spread where they are now is very different percentage wise than a 2% spread when short term rates are 4% or higher.
As per some previous posts, it would seem that when the yield curve is 0% to 2%, that 3 asset classes would be appropriate, that being SP500, EAFE, and GSCI. For yield curves over 2%, then 3 or even 4 classes are good, those being SP500, EAFE, Bonds, and REIT. The logic is the more classes, the better the diversity of the allocation. I would suspect that drawdowns might be lessened by doing this but I can't truly claim this.
For yield curves under 0%, only 2 classes seem possible.
will take a look if I have time…
Yeah, maybe divide by all 4 asset classes when above 2%…makes more sense
ok ok i'll update the post with the most optimal results….
will take a look if I have time…
ok, how do you prefer to look at it….(and I agree in theory)
i'm just surprised no one is mentioning combining the momentum and yield curve systems…..ahem ahem…
That would seem to be tricky since only 2 classes are available to rotate into in the first place when the yield curve is below 0%.
However, when the yield curve is above 2%, you do have a possible choice of 4 classes from which you could rotate into the best 2…
Is the momentum the asset class rotation study?
yah
The curve is between 0-2% 42% of the time and above 2% 45% of the time. Does it increase the total return if you break down these two into smaller increments? Perhaps 0-1%, 1-2%, 2-3% and >3%? Thank you!
Ok, Meb, I'm confused. How did you combine the momentum and yield curve systems? Can you walk us through the mechanics? Thanks
Great work as always Meb. How about using the yield curve model with the TAA model – in other words, buy the yield curve rotation system only if the corresponding asset class index is also above a long-term moving average? I suspect the indices may be above long term moving averages most of the time when the yield curve is giving a long signal for the various asset classes, but it would be interesting to look at.
I try to use a kind of “yield curve” device based on changes in short term interest rates. It behaves much like your blue “yield curve rotation” chart. Note that it stopped being the best performer in late 1999 or so, and was just OK after ~1988 (1988-1999). I think this does have value.
I have been trying to find the reason for this behavior for some time. It appears to be related to the underlying inflation forces. When they get weaker, the yield curve seems to drive the stock market performance less. I haven't done any quantitative work on this, but I have observed what your chart implies.
Thanks for the post.
The results for bonds are distorted by your choice of the 30YT as the vehicle. There might be a position for SHY when the yield curve inverts, but definitely not one for the TLT.
Another “asset class” to look at might be non-marked-to-market bonds, i.e. an income ladder. A 10-year Treasury ladder held to maturity really has NO price movement, per se, but you get 10 years of income. As the purchases mature, the funds are reinvested for income with fresh 10YTs. The initial setup requires buys up and down the curve, though. In the context of a portfolio with non-bonds included, the 10YT ladder could be “rebalanced” by adjusting how much is purchased at the long end each year, to keep the total cash in the ladder equal to the proper proportion of the total.
If you want to look at econometrics for buying junk and corporate instead of Ts, I suggest a mix of multi-year changes in CPI and a non-linear relationship with prevailing 10YT rates. Of course, I’m certain that relative momentum works just as well in marked-to-market fixed income as it does in equities …
The results for bonds are distorted by your choice of the 30YT as the vehicle. There might be a position for SHY when the yield curve inverts, but definitely not one for the TLT.
Another “asset class” to look at might be non-marked-to-market bonds, i.e. an income ladder. A 10-year Treasury ladder held to maturity really has NO price movement, per se, but you get 10 years of income. As the purchases mature, the funds are reinvested for income with fresh 10YTs. The initial setup requires buys up and down the curve, though. In the context of a portfolio with non-bonds included, the 10YT ladder could be “rebalanced” by adjusting how much is purchased at the long end each year, to keep the total cash in the ladder equal to the proper proportion of the total.
If you want to look at econometrics for buying junk and corporate instead of Ts, I suggest a mix of multi-year changes in CPI and a non-linear relationship with prevailing 10YT rates. Of course, I’m certain that relative momentum works just as well in marked-to-market fixed income as it does in equities …
The yield curve is important to watch because it has an influence on multiple asset classes, including stocks.
But with volatility in stocks still high, how is an investor supposed to make good returns? That is where the market timing signals come into the picture and they can enhance an investor’s returns.
Consider http://invetrics.com
Its daily DJIA index trading signal is up a respectable 68% for the year (as of November 1, 2009) and it is free of charge for individual investors.
[...] is an old post where we took at a look at investing based on the yield curve. I updated the charts here with more and less granularity. From the tables one could infer that [...]